Proposed research is to apply computer ultrasound to breast cancer diagnosis. Using special transducers and powerful pattern recognition computer programs we will collect clinical data and search for highly discriminating echo features and classification rules. The major goals are to (1) obtain characteristic features of digitized ultrasound signals scattered from breast tumors in order to (2) determine classification rules and error probabilities for quantitative computer diagnosis and (3) accumulate data base for use in evaluating use of methods in ultrasonic breast cancer screening programs.